vol. 180, no. 1
the american naturalist
july 2012
Did Pollination Shifts Drive Diversification in Southern
African Gladiolus? Evaluating the Model of
Pollinator-Driven Speciation
Luis M. Valente,1,2,* John C. Manning,3 Peter Goldblatt,4 and Pablo Vargas1
1. Real Jardı́n Botánico de Madrid, Consejo Superior de Investigaciones Cientı́ficas, Plaza Murillo 2, 28014 Madrid, Spain; 2. Imperial
College London, Silwood Park Campus, Ascot, Berkshire SL5 7PY, United Kingdom; 3. South African National Biodiversity Institute,
Kirstenbosch, Private Bag X7, Cape Town, South Africa; and Research Centre for Plant Growth and Development, School of Biological
and Conservation Sciences, University of KwaZulu-Natal, Pietermaritzburg, Private Bag X01, Scottsville 3209, South Africa; 4. Missouri
Botanical Garden, St. Louis, Missouri 63166
Submitted August 17, 2011; Accepted March 22, 2012; Electronically published May 24, 2012
Dryad data: http://dx.doi.org/10.5061/dryad.7pr554s8.
abstract: The pollinator-driven ecological speciation model has
frequently been invoked to explain plant richness in biodiversity
hotspots. Here, by focusing on Gladiolus (260 species), a flagship
example of a clade with diverse pollination biology, we test the hypothesis that high species diversity in southern Africa, one of the
world’s most floristically rich regions, has primarily been driven by
ecological shifts in pollination systems. We use phylogenetic methods
to estimate rates of transition between the seven highly specialized
pollination strategies in Gladiolus. We find that pollination systems
have evolved multiple times and that some pollination strategies arose
by a variety of evolutionary pathways. Pollination shifts account for
up to one-third of all lineage splitting events in the genus, providing
partial support for the pollinator-driven speciation model. Transitions from the ancestral pollination mode to derived systems have
also resulted in increased rates of diversification, suggesting that certain pollination systems may speed up speciation processes, independently of pollination shifts per se. This study suggests that frequent pollination shifts have played a role in driving high phenotypic
and species diversity but indicates that additional factors need to be
invoked to account for the spectacular diversification in southern
African Gladiolus.
Keywords: pollination, Gladiolus, Iridaceae, phylogeny, radiation,
speciation.
Introduction
The interaction between plants and their animal pollinators has been hypothesized to play a significant role in
angiosperm diversification (Stebbins 1970; Kay et al.
2006). The study of the evolution of pollination systems
can thus be expected to provide valuable insights into the
* Corresponding author; e-mail: l.valente@imperial.ac.uk.
Am. Nat. 2012. Vol. 180, pp. 83–98. 䉷 2012 by The University of Chicago.
0003-0147/2012/18001-53264$15.00. All rights reserved.
DOI: 10.1086/666003
causes of extant patterns of species richness in biodiversity
hotspots, particularly in clades characterized by specialized
pollination systems (Johnson 2010). The existence of a
causal relationship between diversity of pollination systems
and species richness, however, still remains controversial,
even for regions and clades characterized by an unusual
variety of floral types (Kay and Sargent 2009; Smith 2010).
One of the evolutionary settings where the role of pollination systems as drivers of angiosperm diversification
has been most often invoked is the southern African subcontinent (Johnson 2010). This is a region of exceptional
botanical diversity, harboring more than 20,000 plant species comprising 5%–8% of the world’s total number of
plant species (Goldblatt and Manning 2002). This diversity
is mostly concentrated in the Cape Floristic Region, Succulent Karoo, and the Maputaland-Pondoland-Albany
biodiversity hotspots (Linder 2003; Mittermeier et al.
2005). Several large plant clades in southern Africa exhibit
a high diversity and specialization of pollination systems
(e.g., clades in Amaryllidaceae, Geraniaceae, Iridaceae, Orchidaceae, Scrophulariaceae), leading to the hypothesis
that the radiation of these florally diverse lineages has been
driven by plant-pollinator interactions (Johnson 1996; Van
Der Niet and Johnson 2009). This model has received
much attention in the literature (Johnson 2010) but evolutionary evidence to support it remains sparse. Previous
empirical studies of genera with high species richness in
the Cape region of southern Africa, namely in Orchidaceae
(Johnson et al. 1998; Johnson and Kurzweil 1998; Waterman et al. 2011) and Iridaceae (Goldblatt and Manning
1998), have found a link between plant diversity in the
region and adaptive radiation of pollination systems. The
diverse, albeit scarce, pollinator fauna and a complex geographic mosaic of pollinators in southern Africa have been
84 The American Naturalist
proposed as potential reasons for strong selection for floral
traits that attract pollinators in this region (Johnson 1996;
Barraclough 2006; Linder et al. 2010). However, a recent
study of southern African clades by Schnitzler and colleagues (2011) found that other ecological factors, such as
edaphic preferences, were better predictors of speciation
rates than were pollination systems. More generally, just
five (Disa, Erica, Gladiolus, Pelargonium, and Moraea) of
the 20 largest genera (180 species) in the Cape Floral Region (Goldblatt and Manning 2002) have diversified florally to any extent that is relevant to their pollination
biology, with the great majority of the genera highly conserved in functional floral morphology and thus pollination system. Pollinator-related radiation is only one of
several characteristics of the southern African flora, and
in only a minority of plant groups can it be anticipated
to be the primary driver of speciation.
An important prerequisite for understanding the role
of pollinator shifts in southern Africa is the quantification
of the degree of evolutionary flexibility of pollination systems within the major plant clades in the flora. Evidence
from empirical studies on a variety of southern African
lineages dominated by specialized pollination systems has
revealed that some floral traits associated with different
classes of pollinator can indeed be highly labile (e.g., Johnson et al. 1998; van der Niet and Johnson 2009; Schnitzler
et al. 2011). However, little is known about the frequencies,
directionality, and rates of evolutionary shifts between different pollination systems, not only in southern African
clades but also in angiosperms in general (Perret et al.
2003; Pérez et al. 2006; Bastida et al. 2010; Smith 2010).
If pollinator-mediated divergence is the predominant
model of speciation in southern Africa, then the assumption of high evolutionary flexibility of pollination systems
must be fulfilled.
Here, we test the hypothesis that high species diversity
in southern Africa has been caused by frequent ecological
shifts in pollination system, as proposed by Johnson
(1996), by focusing on Gladiolus, one of the largest genera
in southern Africa and the most striking example of a
clade with diverse pollination biology in the region. Gladiolus (Iridaceae) comprises more than 260 species and
presents one of the widest spectra of pollination systems
known in angiosperms, with at least seven specialized systems described among the southern African species alone
(fig. 1; Goldblatt and Manning 1998). The genus extends
through temperate and tropical Africa, Madagascar, the
Mediterranean basin, and the Middle East, but its main
center of diversification is in southern Africa, where more
than 65% of the species occur.
Gladiolus is an ideal system to evaluate the pollinatordriven ecological speciation model because it forms a
monophyletic assemblage (Valente et al. 2011), and its
species are typically highly specialized for pollination by
a single animal group (Goldblatt et al. 2001), allowing the
frequency and directionality of pollinator shifts to be quantified in a rigorous fashion. In addition, in contrast to
other large genera, the pollination biology of Gladiolus is
well understood, particularly in southern Africa, where
more than half of the species have been subject to intense
field observations of pollination ecology (for a review, see
Goldblatt et al. 2001). Detailed information on types of
animal visitors and pollination mechanisms is available for
more than 80 species. Analysis of the main flower types
in Gladiolus has shown that flower type correlates closely
with pollination system (Goldblatt et al. 2001). As a result,
Goldblatt and Manning (2006) were able to infer, with a
high degree of confidence, the pollination strategies of a
further 133 species based on floral syndromes. Of the 213
species of Gladiolus whose pollination systems are now
known or inferred, more than 50% are pollinated by longtongued bees in the family Apidae that forage primarily
for nectar. This system is the predominant mode of pollination in several species-rich southern African genera of
Iridaceae and is thought to be ancestral in many (Goldblatt
and Manning 2006), raising the question as to whether it
may have led to elevated rates of diversification in these
clades. Other specialized pollination systems recorded in
the genus are, in order of importance, long-proboscid flies
(Nemestrinidiae and Tabanidae), sunbirds (Nectarinia
spp.), night-flying moths (Noctuidae and Sphingidae),
butterflies (Aeropetes tulbaghia), and short-tongued female
pollen-collecting bees (Halictidae and Andrenidae). The
rarest pollination strategy in the genus is by hopliine beetles (Scarabaeidae: Hopliini), which has been confirmed
in only one species.
In this study, we reconstruct the evolutionary history
of pollination biology in Gladiolus, taking advantage of a
recent species-level phylogenetic study of the genus by
Valente et al. (2011), who sampled more than 80% of
southern African species (132 spp.) and all seven Mediterranean basin species. We (i) estimate the degree of lability and also the directionality of pollination system transitions in Gladiolus in order to quantify the possible role
of these ecological shifts in speciation; (ii) test whether
transitions between the predominant pollination system
in the genus—long-tongued bee pollination—and the
rarer pollination systems might have influenced rates of
diversification; and (iii) examine the evolution of pollination strategies in a rapidly diversifying subclade of the
genus endemic to southern Africa in order to gain insights
into the way that ecological shifts operate at lower phylogenetic levels.
A
B
C
D
E
F
G
H
86 The American Naturalist
Methods
Phylogenetic Framework
Phylogenetic analyses were conducted based on the Gladiolus species-level molecular data set of Valente et al.
(2011). This matrix consists of five plastid regions (matK,
psbA-trnH, trnS-trnG, rpl32-trnL, and trnQ-rps16) and includes 148 species of Gladiolus. The five-marker plastid
data set was reanalyzed using both maximum likelihood
and Bayesian phylogenetic methods, following the approach of Valente et al. (2011). However, unlike as in this
previous study, we also added information from indels. A
matrix of indel characters was produced using the “simple
indel coding” method (Simmons and Ochoterena 2000)
as implemented in Seqstate (Müller 2005). In the Bayesian
analyses, the indel character matrix was analyzed under a
binary model implemented in MrBayes 3.1.2 (Ronquist
and Huelsenbeck 2003). For the maximum likelihood
analyses, we used a BINCAT approximation to the indel
data, as implemented in RaxML 7.2.1 (Stamatakis 2006).
Scoring Pollination Systems
Data on Gladiolus pollination systems was extracted from
the following published accounts: Goldblatt 1996; Goldblatt and Manning 1998; and Goldblatt et al. 2001. In
addition, new field observations were conducted for the
Mediterranean basin species Gladiolus communis and
Gladiolus illyricus. Detailed direct field observations are
thus available for 87 species of Gladiolus (available in table
A1 in Dryad: http://dx.doi.org/doi:10.5061/dryad
.7pr554s8).
Pollination system was assigned based on the primary
floral visitor type observed, which we consider to be the
pollinator group following Goldblatt et al. (2001). For all
species of Gladiolus (except for Gladiolus meliusculus; see
below), flowers were visited either by a single animal species or by only one of seven ecologically homologous
groups of species (functional groups) as defined by Goldblatt and Manning (2006). Given that each species of Gladiolus was pollinated only by one animal group, all species
of Gladiolus can be considered to have a specialized pollination system (Fenster et al. 2004). The only exception
was G. meliusculus, which is unusual in the genus in demonstrating a bimodal pollination system by both hopliine
beetles and long-tongued bees (Goldblatt et al. 2001). For
simplicity we refer to this system as “beetle pollination”
from here onward.
Pollination system was inferred for 58 species for which
no field data was available, as it has been shown that flower
type correlates strongly with pollination system in Gladiolus (Goldblatt et al. 2001), and there is a strong correspondence between pollination system assigned and actual
observations (Goldblatt et al. 2001). For three species (G.
decoratus and G. erectiflorus from tropical Africa; and G.
horombensis from Madagascar) assignation of pollination
system based on floral syndrome was not possible, and the
character was coded as unknown.
The pollination systems of each of the 148 species sampled and the method of pollination system scoring are
shown in table A1 (available in Dryad: http://dx.doi.org/
doi:10.5061/dryad.7pr554s8). Each species was assigned
one of the following character states, based on the classification of specialized pollination systems proposed by
Goldblatt et al. (2001): pollination by long-tongued apid
and anthophorine bees that forage for nectar (LB); pollination by short-tongued halictid and andrenid bees that
collect pollen (SB); passerine bird pollination (BI); nightflying moth pollination (MO); satyriine butterfly pollination (BU); hopliine beetle pollination (BE); long-proboscid fly pollination (LF); unknown (?).
Phylogenetic Signal of Pollination System
We evaluated the phylogenetic signal of pollination system
using a parsimony approach in Mesquite, version 2.72
(Maddison and Maddison 2009). Taxa were randomly reshuffled among tips of each of 1,000 trees from the
MrBayes output and the number of parsimony steps required to explain the evolution of the trait in each of the
new trees was counted, thus generating a null distribution
of character steps. We then calculated the average number
of parsimony steps required to explain evolution of the
character in the original 1,000 Bayesian trees without reshuffled terminal taxa. If this value was lower than the
lower 5% percentile of the null distribution, this was considered as evidence for significant phylogenetic signal in
pollination syndrome.
Ancestral Pollination Systems
We reconstructed ancestral character states for pollination
systems using a maximum likelihood method implemented in Mesquite, version 2.72. To account for uncertainty in tree topology and branch lengths, character optimizations were repeated for each of 1,000 highly probable
Figure 1: Diversity of pollination systems in Gladiolus; species shown and their pollination system: A, G. saundersii, butterfly; B, G. meliusculus,
beetle; C, G. reginae, long-proboscid fly; D, G. carinatus, long-tongued bee; E, G. dalenii, sunbird; F, G. atroviolaceus, long-tongued bee; G,
G. quadrangulus, short-tongued bee; H, G. tristis, moth. All species shown are from southern Africa, except G. atroviolaceus from Eurasia.
Pollination Systems in Gladiolus
trees from the MrBayes output, excluding the outgroup.
We selected the single-parameter Markov k-state model,
thus assigning equal probability to any type of pollination
system shift. The best estimate of character state for each
node in each of the trees was determined using the likelihood ratio test with a decision threshold of two, which
was chosen in order to make a conservative assignation
of character states at nodes (Maddison and Maddison
2009). Reconstructions were considered equivocal if the
difference in log likelihood between alternative reconstructions was below the threshold.
Pollination System Shifts: Frequency and Rates
To estimate the frequency of shifts between each of the
seven pollination systems we used the “summarize state
changes over trees” application in Mesquite. The number
of transitions between different character states was
counted in each of the 1,000 resampled Bayesian trees
using both maximum likelihood and unordered parsimony
reconstruction methods.
The relative rates of transition between pollination systems were estimated using SIMMAP, version 15, build
26022010.1 (Bollback 2006). This software implements a
Bayesian stochastic character mapping method (Huelsenbeck et al. 2003) that reconstructs character states and
transitions based on a user-specified prior probability distribution of the rate of change. We performed two sets of
SIMMAP analyses. First, in order to model the evolution
of the seven pollination systems, we coded each system as
a unique character state. Second, in order to explicitly
investigate the rates of change from the most common
pollination system—long-tongued bee—to all six remaining systems (as well as in the opposite direction), we coded
pollination system as a binary character. In order to choose
suitable prior parameters we followed a two-step statistical
approach, as recommended for SIMMAP, version 1.5.
First, we performed a Markov chain Monte Carlo analysis
(1 # 105 cycles, sampling every 200 cycles) on the best
tree from MrBayes to sample overall rate values (for both
seven-state and binary character analyses) and bias values
(binary character analysis only). We then used the posterior distribution of these parameters as the input for an
R script that finds the parameter values offering the best
fitting gamma and beta distributions (Bollback 2006; available from www.simmap.com). We repeated this procedure
using 10 burn-in optimal trees randomly chosen from the
MrBayes output and found that although optimal parameter values varied slightly, this did not strongly affect the
results on relative rates of state change. We therefore report
the results obtained using the parameter values estimated
in the analysis of the MrBayes consensus tree. These were
the following: seven-state analysis a p 6.43, b p 0.013;
87
two-state analysis a p 7.46, b p 0.024, a (beta
distribution) p 48.18. In addition, for the seven-state
analysis we assigned an empirical prior on state frequencies. In order to incorporate phylogenetic uncertainty in
our data set, analyses were performed on 1,000 Bayesian
trees (after pruning the outgroup). Ten realizations were
sampled from the prior distributions and 10 realizations
were sampled from each tree, totalling 100,000 mappings.
The analyses yielded posterior distributions for the expected number of transitions between any given state and
for the expected frequency of time spent in a particular
state. We calculated relative rates of state transition for
each realization using the following equation: E[i r
j]/{E[time(i)] # E[rate]}, where E(i r j) is the expected total number of transitions from state i to state j,
E[time(i)] is the expected dwell time in each state and
E[rate] is the mean gamma rate (J. Bollback, personal
communication). This allowed us to obtain a posterior
distribution of relative expected rates of transition for all
combinations of i and j.
Character-Dependent Speciation and Extinction Rates
In order to investigate whether given types of pollination
system have promoted diversification in Gladiolus, we used
the Binary State Speciation and Extinction (BiSSE) model
(Maddison et al. 2007; FitzJohn et al. 2009) implemented
in the R package diversitree (FitzJohn et al. 2009). This
method does not test the pollination shift speciation hypothesis, given that it does not assess whether higher numbers of shifts lead to higher rates of speciation. It estimates
binary trait-dependent speciation and extinction rates in
a Bayesian framework. We use BiSSE here solely to evaluate
whether there is a character-dependent effect on speciation
or extinction rates. A multistate extension of BiSSE
(MuSSE) is also available, but we did not use it as the
character we are interested in has seven states, which would
require the model to estimate 56 parameters and would
limit the interpretation of results (R. FitzJohn, personal
communication). We therefore restrict our analyses to the
comparison of the effect on diversification rate of the predominant pollination system (long-tongued bee; see table
A1, available in Dryad: http://dx.doi.org/doi:10.5061/
dryad.7pr554s8) versus all other systems (rare systems).
We did not run similar models for other types of pollination systems (e.g., moth pollination versus all other systems) because the number of species with each of the
derived systems is small, and this would reduce the accuracy of parameter estimation (Maddison et al. 2007).
Given that our species sample is random with respect to
pollination system, we assume that the number of species
with each system included in this study is proportional to
the total number of species that exist with that given char-
88 The American Naturalist
/
▲/▲
/
/▲
/
/
/
Long-tongued bee
/
Moth
/
/
Short-tongued bee
Sunbird
Long-proboscid fly
/
/
/▲
/
/
▲/
Butterfly
Beetle
Equivocal reconstruction
Node absent
/
/
/
/
/
/
/
/
/
/
/
/
/
/
/
/▲
/
/
/
/▲
/
/
/
/
/
/
/
/
/
/
Cape radiation
clade
Continued in Fig. 2B
acter state. In order to obtain phylogenetic trees that are
proportional to time, we used the uncorrelated lognormal
divergence dating method implemented in BEAST (Drummond and Rambaut 2007), fixing the stem age of Gladiolus
at 25.8 million years following Valente et al. (2011). Note
that we are interested only in the relative rates of diversification, and therefore the absolute divergence time estimates are not relevant. In diversitree, we ran two independent MCMC chains for 10 # 10 4 steps for each of 10
G. marlothii
G. mostertiae
G. viridiflorus
G. salteri
G. deserticola
G. arcuatus
G. orchidiflorus
G. lapeirousioides
G. scullyi
G. venustus
G. virescens
G. uysiae
G. inflatus
G. cylindraceus
G. quadrangularis
G. oreocharis
G. aquamontanus
G. brevitubus
G. subcaerulus
G. bullatus
G. geardii
G. ceresianus
G. longicollis
G. abbreviatus
G. fourcadei
G. huttonii
G. hyalinus
G. liliaceus
G. overbergensis
G. teretifolius
G. tristis
G. acuminatus
G. permeabilis
G. saccatus
G. splendens
G. cunonius
G. vandermerwei
G. taubertianus
G. pritzelii
G. insolens
G. pulcherrimus
G. equitans
G. alatus
G. meliusculus
G. speciosus
G. bilineatus
G. stefaniae
G. carmineus
G. rudis
G. floribundus
G. grandiflorus
G. miniatus
G. buckerveldii
G. albens
G. vaginatus
G. variegatus
G. carneus
G. angustus
G. undulatus
G. debilis
G. emiliae
G. guthriei
G. blommesteinii
G. comptonii
G. aureus
G. bonaspei
G. brevifolius
G. caerulus
G. carinatus
G. caryophyllaceus
G. engysiphon
G. exilis
G. gracilis
G. griseus
G. hirsutus
G. inflexus
G. jonquilliodorus
G. maculatus
G. martlyei
G. meridionalis
G. monticola
G. nerineoides
G. ornatus
G. priorii
G. quadrangulus
G. recurvus
G. rogersii
G. trichonemifolius
G. vigilans
G. virgatus
randomly chosen trees using an exponential distribution
prior for the rates. We discarded the first 25% of steps of
each chain as burn-in.
The Cape Radiation Clade
A radiation involves proliferation of species over a short
period of time from an initial ancestor and may reflect
the potential role of pollinator shifts in the long term. We
Pollination Systems in Gladiolus
/
Continued
from Fig. 2A
Mediterranean
basin species
/
/
/
▲/
/
/
▲/▲
▲/
▲/
▲/
/▲
/
/
/▲
/
▲/
/
/▲
/
/▲
/▲
89
G. horombensis
G. triphyllus
G. gregarius
G. anatolicus
G. atroviolaceus
G. italicus
G. communis
G. illyricus
G. unguiculatus
G. imbricatus
G. candidus
G. aequinoctialis
G. decoratus
G. gracillimus
G. robiliartianus
G. oatesii
G. pretoriensis
G. loteniensis
G. erectiflorus
G. parvulus
G. gueinzii
G. inandensis
G. wilsonii
G. involutus
G. stellatus
G. cruentus
G. papilio
G. pardalinus
G. malvinus
G. woodii
G. mortonius
G. ochroleucus
G. aurantiacus
G. crassifolius
G. dolomiticus
G. flanaganii
G. kamiesbergensis
G. oppositiflorus
G. reginae
G. saundersii
G. sericeovillosus
G. hollandii
G. serpenticola
G. ecklonii
G. exiguus
G. macneilii
G. pole evansii
G. rufomarginatus
G. vinosomaculatus
G. densiflorus
G. elliotii
G. pavonia
G. dalenii
G. calcaratus
G. ferrugineus
G. appendiculatus
G. varius
Figure 2: Evolution of pollination systems in Gladiolus. The tree shown is the majority rule consensus from the MrBayes (Ronquist and
Huelsenbeck 2003) analysis. Symbols above the nodes represent Bayesian posterior probabilities/maximum likelihood bootstrap values.
Lozenges represent posterior probabilities above 95% and bootstrap values above 80%. Triangles represent posterior probabilities above
90% and bootstrap values above 60%. Pie charts at selected nodes summarize the results of the maximum likelihood character optimization
analyses in Mesquite (Maddison and Maddison 2009), conducted for 1,000 highly probable Bayesian trees. Each chart shows the percentage
of trees for which a given pollination system was reconstructed as ancestral for that node. Branches are colored according to a parsimony
mapping of ancestral pollination systems conducted in Mesquite onto the tree shown in the figure. The crown nodes of the Cape radiation
clade and the Mediterranean basin lineage are indicated by arrows.
provide new molecular data for a lineage of 28 species of
Gladiolus endemic to the southwesternmost region of the
Cape of southern Africa. Despite an apparently recent origin (1.5–3.0 million years; Valente et al. 2011), this lineage
is strikingly diverse in terms of floral morphology, with
six different specialized pollination systems confirmed.
Due to low sequence divergence, however, the molecular
matrix of Valente et al. (2011) could not resolve relationships among the species of this clade, which they named
the “Cape explosion” (Cape radiation clade, fig. 2). We
therefore expanded the sampling of the genome for these
taxa, producing new sequences for three additional plastid
regions (petL-psbE, psbJ-petA, and ndhJ-trnF) for the 28
species that comprise the radiation (Genbank accession
numbers JQ796185–JQ796273). We investigated relationships among plastid haplotypes within 27 of the 28 species
(we excluded Gladiolus caryophyllaceus as we were unable
to generate sequences for two of the eight markers). The
eight plastid regions were concatenated and the resulting
matrix was analyzed using a coalescence-based method
implemented in TCS (Clement et al. 2000). A haplotype
network was constructed, with gaps treated as missing data
and with a connection limit of 95%.
We used Bayes factors (BFs) to test whether different
90 The American Naturalist
pollination systems have evolved multiple times within
the Cape radiation clade. The likelihood obtained under
an unconstrained Bayesian inference analysis was compared with that obtained under analyses with the different pollination systems constrained to be monophyletic.
This test was not conducted for butterfly pollination because there is only one species within the Cape radiation
clade with this pollination system. BFs were calculated
using Tracer, and a monophyly hypothesis was rejected
if ⫺2 log (BF) 1 10.
Results
Phylogenetic Analyses
The topology retrieved in the reanalysis of the Gladiolus
matrix from Valente et al. (2011) was almost identical to
that obtained in the previous study, except for differences
at certain terminal nodes, reflecting low variation of sequences rather than strongly supported conflict (phylogenetic tree available in Dryad: http://dx.doi.org/doi:
10.5061/dryad.7pr554s8) The addition of the indel character matrix led to an increase in both posterior probabilities and bootstrap support values for the majority of
nodes. The maximum likelihood topology recovered in the
RaxML analysis was fully congruent with that obtained in
the majority-rule consensus tree in MrBayes, and we therefore used the trees obtained by Bayesian inference for all
subsequent analyses.
Evolutionary History of Pollination Systems
Pollination system exhibited a significant level of phylogenetic signal according to the analysis integrating over
several topologies in Mesquite. The average number of
parsimony steps required to explain the evolution of pollination systems in Gladiolus was 50.6 (mean number of
steps in 1,000 trees). This value is less than the 0.05 percentile threshold (56 steps) of the distribution of character
steps generated by reshuffling taxa in 1,000 randomizations. Therefore, closely related species tend to share pollination systems more often than expected by chance.
The results of the maximum-likelihood optimizations
of pollination system evolution conducted more than
1,000 Bayesian trees are summarized in figure 2. Longtongued bee pollination was unambiguously reconstructed
as the ancestral state for the root of Gladiolus in 99.3% of
trees.
The average, minimum, and maximum number of gains
and losses of each pollination system in the 1,000 trees
analyzed are shown in table 1. The differences found between the parsimony and maximum likelihood results
stem from the fact that our choice of a conservative like-
Table 1: Number of unequivocal independent gains and losses
of each of the pollination systems throughout the evolutionary
history of Gladiolus
Parsimony
ML
Average Min Max Average Min Max
Total shifts
Independent gains:
Long-tongued bee
Moth
Short-tongued bee
Sunbird
Beetle
Long-proboscid fly
Butterfly
Independent losses:
Long-tongued bee
Moth
Short-tongued bee
Sunbird
Beetle
Long-proboscid fly
Butterfly
51.4
48
53
27.5
14
40
3.4
7.9
4.9
11.4
1.0
17.1
5.7
1a
5
4
8
1
12
4
8
10
5
14
1
20
6
.4
3.4
4.2
5.7
.9
9.5
3.5
1a
1a
2
2
1a
4
1
3
8
5
10
1
15
6
38.4
2.4
.2
3.1
.0
6.3
1.0
28
0
0
0
0
0
0
47
10
5
11
0
14
8
24.9
.4
.0
1.5
.0
.6
.2
14
0
0
0
0
0
0
38
4
2
5
1
6
3
Note: As inferred in the character optimization analyses conducted in Mesquite more than 1,000 highly probable trees, using parsimony and maximum
likelihood (ML) methods.
a
Minimum gains were set to 1 in the cases where Mesquite could not infer
any unequivocal shifts toward a particular state.
lihood threshold resulted in more equivocal assignations
of character states at nodes in the likelihood method, and
thus fewer transitions in the likelihood analysis.
Overall, a minimum of 48 (parsimony) or 14 (maximum likelihood [ML]) pollination shifts was inferred. The
pollination system that was reconstructed to have been
gained most frequently is long-proboscid fly pollination,
with a minimum of 12 (parsimony) or four (ML) independent gains. Bird pollination also has a high number of
origins, having evolved a minimum of eight (parsimony)
or two (ML) times. Moth, butterfly, and short-tongued
bee pollination systems have all emerged more than once
as revealed by both methods. Long-tongued bee pollination has been lost a minimum of 28 (parsimony) or 14
(ML) times. According to most trees, all pollination systems, with the exception of short-tongued bee and beetle
pollination, have each been lost at least once (table 1).
Character optimization analyses rendered similar results
using parsimony, maximum likelihood, and Bayesian stochastic approaches (table 2). Of the 42 possible types of
character state transitions, the one that has occurred most
often has been from pollination by long-tongued bee to
long-proboscid fly. Transitions from long-tongued bee pollination to moth, short-tongued bee, bird, and butterfly
pollination have all occurred more than once.
Table 2: Frequency of each of the 42 types of possible pollination system transitions in Gladiolus
Transition
Parsimony (Mesquite)
ML (Mesquite)
Bayesian stochastic
mapping (SIMMAP)
LBrMO
LBrSB
LBrBI
LBrBE
LBrLF
LBrBU
MOrLB
MOrSB
MOrBI
MOrBE
MOrLF
MOrBU
SBrLB
SBrMO
SBrBI
SBrBE
SBrLF
SBrBU
BIrLB
BIrMO
BIrSB
BIrBE
BIrLF
BIrBU
BErLB
BErMO
BErSB
BErBI
BErLF
BErBU
LFrLB
LFrMO
LFrSB
LFrBI
LFrBE
LFrBU
BUrLB
BUrMO
BUrSB
BUrBI
BUrBE
BUrLF
4.4 (1–7)
4.7 (3–5)
8.2 (5–11)
1.0 (1)
15.5 (11–20)
4.7 (2–6)
.7 (0–4)
.1 (0–1)
1.4 (0–5)
0 (0)
.2 (0–2)
.03 (0–1)
.03 (0–1)
.1 (0–2)
.1 (0–2)
0 (0)
.1 (0–1)
.02 (0–1)
.2 (0–2)
1.9 (0–5)
.1 (0–2)
0 (0)
.8 (0–4)
.1 (0–2)
0 (0)
0 (0)
0 (0)
0 (0)
0 (0)
0 (0)
2.2 (0–7)
1.6 (0–4)
.1 (0–1)
1.6 (0–5)
0 (0)
1.1 (0–4)
.3 (0–3)
.01 (0–1)
.02 (0–1)
.1 (0–2)
0 (0)
.6 (0–4)
2.0 (0–5)
4.1 (2–5)
5.3 (1–8)
.9 (0–1)
9.2 (4–15)
3.4 (1–6)
.2 (0–3)
.02 (0–1)
.2 (0–3)
0 (0)
0.1 (0–1)
0 (0–1)
0 (0–1)
0 (0–1)
0 (0–1)
0 (0)
0 (0–1)
0 (0)
.1 (0–1)
1.3 (0–4)
.02 (0–1)
0 (0)
.1 (0–2)
.02 (0–1)
0 (0–1)
0 (0)
0 (0)
0 (0)
0 (0)
0 (0)
.2 (0–3)
.1 (0–2)
.04 (0–1)
.3 (0–3)
0 (0)
.02 (0–1)
0 (0–1)
0 (0–1)
0 (0–1)
.01 (0–1)
0 (0)
.1 (0–2)
5.7 (0–16)
5.0 (0–12)
10.8 (1–23)
1.2 (0–5)
18.0 (5–33)
4.8 (0–14)
4.1 (0–22)
.2 (0–4)
1.7 (0–10)
.01 (0–3)
1.5 (0–11)
.2 (0–5)
1.6 (0–16)
.1 (0–5)
.2 (0–6)
.01 (0–2)
.3 (0–6)
.09 (0–4)
6.1 (0–27)
2.4 (0–11)
.4 (0–5)
.03 (0–3)
2.0 (0–13)
.5 (0–5)
.3 (0–10)
.02 (0–3)
.01(0–2)
.04 (0–4)
.06 (0–4)
.01 (0–4)
13.1 (0–36)
2.7 (0–11)
.6 (0–7)
2.8 (0–12)
.06 (0–3)
1.2 (0–8)
2.9 (0–17)
.3 (0–5)
.1 (0–4)
.7 (0–7)
.01 (0–2)
1.7 (0–10)
Note: Two columns show the average number of unequivocal transitions between each of the
pollination systems as inferred in the character optimization analyses conducted in Mesquite
(Maddison and Maddison 2009) more than 1,000 highly probable trees, using parsimony and
maximum likelihood (ML) methods. One other column shows the expected number of transitions,
obtained in the Bayesian stochastic mapping analysis of 1,000 highly probable trees in SIMMAP
(Bollback 2006). Numbers in parentheses are the minimum and maximum number of transitions
inferred. LB: long-tongued bee; MO: moth; SB: short-tongued bee; BI: sunbird; BE: beetle; LF:
long-proboscid fly; BU: butterfly.
92 The American Naturalist
Rates of Pollination System Shifts
The posterior distributions of expected relative rates of
transition obtained in the seven-state Bayesian analysis in
SIMMAP are shown in figure 3. On average, the transitions
that occurred at the highest relative rates were the following: long-proboscid fly to long-tongued bee (1.35, 95%
credibility interval [CI] 0.73–2.07), butterfly to longtongued bee (1.28, 95% CI 0–3.04), short-tongued bee to
long-tongued bee (1.19, 95% CI 0–3.35), moth to longtongued bee (1.16, 95% CI 0.24–2.38), and sunbird to
long-tongued bee (0.95, 95% CI 0.25–1.76).
The analyses where pollination was coded as a binary
character (long-tongued bee/non-long-tongued bee pollination) revealed that the transition from long-tongued
bee to other pollination systems occurred at a similar rate
(1.60, 95% CI 1.23–2.03) as the transition in the opposite
direction (1.70, 95% CI 1.28–2.20).
Character-Dependent Speciation and Extinction Rates
The Bayesian BiSSE analyses revealed that both rates of
speciation (l0 and l1) and rates of diversification (speciation minus extinction; r0 and r1) were significantly higher
in Gladiolus lineages with rare pollination systems than in
lineages with the predominant bee-pollinated system (table
3; fig. 4). There was no significant difference between the
rates of transition between the two states (q01 and q10), in
agreement with the SIMMAP analyses.
The Cape Radiation Clade
The new matrix for the Cape radiation clade now includes
a total of eight molecular regions with 7,060 sites. Of the
35 variable sites, 17 (48.5%) were located within the three
new plastid regions sequenced in this study (petL-psbE,
psbJ-petA, ndhJ-trnF). The resulting 19 haplotypes were
connected through other 19 missing haplotypes (extinct
or not found) in a TCS network with no loops (i.e., no
homoplasy) and distributed in two “starlike” clusters (fig.
5). The most common, interior haplotype was shared by
eight species pollinated by four different animal groups
(long-tongued bee, short-tongued bee, moth, and butterfly). Pollination systems did not form independent clusters, and no haplotype was exclusive to a single pollination
system. Instead, species sharing pollination strategies were
mostly found scattered throughout the haplotype network.
Monophyly of long-tongued bee and moth pollination
was strongly rejected in the Bayes factors analysis (table
4), suggesting that these systems have either evolved more
than once within the Cape radiation clade or have evolved
only once and are paraphyletic. However, the monophyly
of short-tongued bee, long-tongued fly, and sunbird pol-
lination could not be rejected, revealing that they only
evolved once within this clade.
Discussion
High Lability of Pollination Systems
The evolution of pollination systems in Gladiolus has been
remarkably dynamic. All pollination systems, with the exception of hopliine beetle pollination, have evolved multiple times. The lability of pollination systems in Gladiolus
was accompanied by unusual versatility in the direction
of shifts: species sharing the same pollination system have,
in some cases, arisen independently from ancestors with
different systems. Long-tongued bee pollination was reconstructed as the ancestral state for Gladiolus, supporting
the untested deduction by Goldblatt and Manning (1998).
The loss of pollination by nectar-collecting bees accounts
for more than 70% of the pollination shifts that have taken
place throughout the radiation of the genus, and the rate
of loss of this system was roughly equal to all combined
rates of transition in the opposite direction (table 3). The
derived pollination strategies were also unusually versatile:
the five highest rates of transition were those that took
place from “derived” pollination systems back to the ancestral long-tongued bee pollination system (fig. 3). The
flexibility of pollination strategies in Gladiolus rivals that
encountered in the genera with the most dynamic histories
of pollination system evolution documented to date, including southern African Disa (Johnson et al. 1998) and
Babiana (Schnitzler et al. 2011) as well as New World
Aquilegia (Whittall and Hodges 2007), Calochortus (Patterson and Givnish 2004), Costus (Kay et al. 2005), Iochroma (Smith et al. 2008), and Ruellia (Tripp and Manos
2008).
Most studies on character evolution tend to find that
although there is often a high diversity of character states,
only a few of the possible types of transitions that could
occur have actually taken place (Whittal and Hodges
2007). In Gladiolus, the percentage of realized character
transitions was also unusually high, with at least 12 of the
42 possible types of pollination system transitions occurring (table 2). Certain specialized pollination strategies in
Gladiolus have evolved from more than one ancestral system; for example, sunbird pollination has evolved independently from moth-, long-tongued bee- and long-proboscid fly-pollinated ancestors (table 2; fig. 3).
Furthermore, at least five of the pollination systems have
been secondarily ancestral to other systems in the genus
(table 1). As far as we know, such lability of pollination
systems has never been documented in any other group.
The seemingly unconstrained directionality of transitions
found in Gladiolus is rare in plants, not only between
Pollination Systems in Gladiolus
LB
To:
From:
10x10
MO
SB
BI
BE
LF
BU
0.16
0.15
0.31
0.03
0.52
0.14
93
5
LB
0
0.2
0
10x10
0.4
0
0.1
0.2
0.3 0.4 0
0.2
0.4
0.6
0
0.05
0.10
0
0.15 0.2 0.4 0.6 0.8 1.0
0.1 0.2 0.3 0.4
5
1.16
MO
0
10x10
1
2
3
0.06
4
0
5
1.19
0
0
1.5
1.0
2.0
0
3.0
0.4
0.8
1.0
0
2.0
3.0
0
0.5
1.0
1.5
2.0
2
4
6
0.1
8
12
0
1
2
3
0.18
4
5
6
0
5
0.01
15 0
10
1.0
0.06
0.25
2.0
3
0
2
4
6
8
10 0
1
2
3
4
5
5
0.95
BI
0
10x10
1.0
0.07
0.42
5
SB
10x10
0.5
0.01
0.49
1.0
2.0
0.39
3.0
0
0.5
0.0
0.06
1.0
1.5
0.4
0
0
0.8
0.1
0.2
0.08
0.32
0.4 0
0.3
0.5
1.0
0
1.5
0.2 0.4 0.6 0.8
5
BE
0.0
0
0.5
1.0
0.0
1.5
2
0.5
1.0
0.0
1.5
2 0
0.5
1.0
0.0
2 0
1.5
0.5
1.0
0.0
1.5
0
2
0.5
1.0
0.0
1.5
2 0
0.5
1.0
1.5
2
5
10x10
LF
1.35
0
10x10
1.0
2.0
3.0
0
0.4
0.3
0.07
0.29
0.8
1.2 0
0.2
0.6 0.8 0
0.4
0.4
0.8
0.01
1.2
0
0.1
0.2
0.12
0.3
0
0.4
0.8
5
BU
0.11
1.28
0
2
4
6
8 10 0
1
2
3
0.3
0.06
4
5 0
1
2
3
4
0
1
2
3
4
0.73
0.01
5
6 0
1
2
3
0
2
4
6
8
Figure 3: Posterior distributions of the expected relative rates of transition between the seven pollination systems in Gladiolus, obtained
in the Bayesian stochastic character mapping analyses conducted more than 1,000 highly probable trees in SIMMAP (Bollback 2006). Each
histogram refers to the transition from the pollination system shown at the start of the row to the pollination system shown at the top the
column. The pollination systems are (in order of appearance from top to bottom in rows and from left to right in columns) LB (longtongued bee), MO (moth), SB (short-tongued bee), BI (sunbird), BE (beetle), LF (long-proboscid fly), and BU (butterfly). The numbers
above each histogram are the averages of the posterior distributions.
pollination states (Tripp and Manos 2008) but also between character states in other types of traits whose transitions also involve major morphological changes (e.g.,
sexual systems; Renner et al. 2009; leaf shape, Jones et al.
2009).
The analysis of the rapidly diversifying Cape radiation
clade illustrates how dynamic pollination system evolution
has been in the genus. The haplotype network revealed
that radiation of pollination systems has proceeded remarkably rapidly in this clade (fig. 5). The six specialized
pollination syndromes that occur within this lineage have
probably all evolved in the last 1.5–3.0 million years (Valente et al. 2011), but little genetic variation has accumulated in the molecular regions sampled. The six pollination strategies did not cluster in distinct lineages within
the haplotype network, indicating that rapid parallel shifts
have occurred. A contrasting scenario was encountered in
the lineage comprising all seven species of Gladiolus from
the Mediterranean basin (fig. 2): speciation in this region
has clearly not been associated with diversification of pol-
94 The American Naturalist
Table 3: Mean and 95% credibility intervals
(CIs) of the parameters estimated in the Bayesian binary state speciation and extinction
analyses
Parameter
Mean
95% CI
l0
l1
m0
m1
r0
r1
q01
q10
.160
.794
.262
.218
⫺.102
.576
.543
1.034
.024–.360
.487–1.178
.017–.633
.007–.684
⫺.455–.151
.269–.899
.097–1.265
.398–2.047
Note: Summary statistics are shown for all Markov
chain Monte Carlo runs combined (each run was conducted independently for each of 20 Bayesian trees). l0
p speciation rate in state 0; l1 p speciation rate in
state 1; m0 p extinction rate in state 0; m1 p extinction
rate in state 1; r0 p diversification rate in state 0; r1 p
diversification rate in state 1; q01 p transition rate from
state 0 to 1; q10 p transition rate from state 1 to 0;
where state 0 is long-tongued bee pollination and state
1 is all other pollination systems combined.
lination systems, as we found that no pollinator shifts have
taken place within this clade (all species are melittophilous), even though the Mediterranean lineage is older (14
million years old; Valente et al. 2011) than the Cape radiation clade. A lower pollinator richness of Europe has
previously been invoked to explain the narrow array of
pollination strategies in this region in Aquilegia (Bastida
et al. 2010), but whether this hypothesis holds for Gladiolus
remains to be tested.
Our estimates of evolutionary flexibility of pollination
systems rely on the topology and branch lengths of the
phylogenetic trees presented in this study. The uncertainty
associated with the lack of support of some branches is a
common feature of phylogenetic species-level studies of
recently diverged groups (Bakker et al. 2005). In all posttree-reconstruction phylogenetic analyses we have explicitly taken this limitation into account by integrating parameter estimations over a large set of Bayesian trees that
capture the uncertainty in the relationships amongst species of Gladiolus. The fact that the Gladiolus phylogenetic
trees used in this study are based solely on plastid data
could mask additional sources of diversity, particularly hybridization, and associated plastid DNA capture, masking
deeper relationships that could be revealed if nuclear data
were available. However, the topology presented here receives external support from the fact that most of the
clades retrieved in our analysis correspond broadly to the
morphology-based series or species groups proposed by
Goldblatt and Manning (1998) in their monograph of the
genus, rather than to geography (Valente et al. 2011), as
would be expected if plastid DNA capture had not been
prevalent.
Pollination Shifts and Speciation in Southern Africa
The complexity of the evolutionary history of pollination
systems in Gladiolus is also illustrated by the high number
of lineage-splitting events that have been associated with
a pollination shift, accounting for between 18.7% and
34.9% of all branching events (average of 27.5 shifts with
maximum likelihood and 51.4 shifts with parsimony). The
fact that an important percentage of branching events in
the genus are linked to pollinator divergence suggests that
pollinator-driven speciation may have been frequent. Nevertheless, our study revealed that the majority of speciation
events in Gladiolus were not associated with a transition
between major pollinator functional groups, therefore providing only partial support for the pollinator-driven speciation hypothesis (Johnson 1996). Indeed, our analysis of
phylogenetic signal of pollination system confirms that,
overall, pollination systems tend to be conserved in Gladiolus and that closely related species tend to share pollination strategy (Pagel 1999). In addition, the Bayes factor
analysis (table 4) revealed that single origin of at least four
(butterfly, long-proboscid fly, short-tongued bee, sunbird)
of the six pollination systems found within the rapidly
evolving Cape radiation cannot be rejected within this lineage. This suggests that although shifts between pollinator
functional groups may have partly driven speciation in
this endemic southern African clade, an additional explanation is required to account for its unusually rapid rates
of diversification (Valente et al 2011), given that most
speciation events were not accompanied by a change in
pollinator category.
Overall, our results show that pollinator shifts have
played a significant role in driving phenotypic diversity
and possibly speciation in one of the most florally diverse
genera. However, it seems that diversification in Gladiolus
follows the pollinator-driven speciation model only to
some extent (Johnson 1996; van der Niet and Johnson
2009), as more than half of the branching events in the
phylogenetic tree appear to be associated with factors other
than pollination biology. This result is particularly relevant
given that it was obtained from the prime example of a
southern African clade exhibiting high diversity of floral
types and pollination strategies (Goldblatt and Manning
2002). This work joins other studies in demonstrating only
a moderate link between speciation and pollinator divergence in the southern Africa flora. Previous analyses of
Cape clades Lapeirousia (Goldblatt and Manning 1996),
Babiana, Moraea, Podalyrieae, and Protea (Schnitzler et al.
2011), have found that pollinator systems show an unexpectedly high degree of conservatism and have alter-
Pollination Systems in Gladiolus
Probability density
5
Ȝ0
Ȝ1
4
ȝ0
ȝ1
3
3
2
2
1
1
0
0
0.0
0.5
1.0
1.5
0.0
0.2
Speciation rate
0.4
0.6
0.8
1.0
1.2
Extinction rate
2.5
Probability density
95
q 01
q 10
r0
r1
2.0
1.0
1.5
1.0
0.5
0.5
0.0
0.0
-1.0
-0.5
0.0
0.5
1.0
Diversification rate
0
1
2
3
4
Character transition rate
Figure 4: Posterior distributions of parameters estimated in the Bayesian binary state speciation and extinction analyses. Horizontal bars
indicate the 95% credibility interval for each parameter. Abbreviations according to table 3.
natively proposed a primary role of edaphic factors in
driving divergence (see van der Niet et al 2006). We are
currently unable to test the edaphic shifts hypothesis in
Gladiolus due to the lack of available relevant data for such
a large group, which has been little studied in an important
part of its range outside of southern Africa. Johnson (2010)
has recently proposed that a link between pollinator distribution and habitat type would facilitate parallel shifts
in pollination system and edaphic specialization. Operation of this model requires that pollination systems are
composed of single pollinator species, in order that shifts
in habitat types become shifts out of an existing pollination
ambit. This has not been demonstrated to apply in Gladiolus, nor in other southern African Iridaceae, in which
the majority of pollinators are functional groups rather
than single species, or have foraging ranges that extend
well beyond individual habitat boundaries.
In this study we have restricted our focus to shifts between the seven pollinator functional categories in Gladiolus. As a result, transitions that may have taken place
within these groups—for example, between different types
of long-proboscid flies or moths—and that may have been
related to speciation could not be assessed. Previous studies have shown that these types of shifts have occurred
between sister species in other southern African clades,
namely Diseae orchids (Johnson and Steiner 1997; Johnson
et al. 1998; Waterman et al 2011). In Gladiolus, a study of
selection by Anderson et al (2009) found evidence for the
buildup of isolating barriers between two floral morphs
of the southern African species G. longicollis that differ in
corolla tube length as a result of adaptation to two hawkmoth guilds. However, a follow-up genetic modeling analysis of this system found that divergence toward speciation
was unlikely (Rymer et al 2011). Although there is currently no evidence to suggest that the Gladiolus pollinator
functional groups considered in this study may conceal
transitions that were relevant to the speciation process, we
cannot exclude this possibility.
In addition to evaluating the pollinator shift speciation
hypothesis, we have assessed whether different types of
96 The American Naturalist
G. virgatus
G. inflexus
G. bonaspei
G. rogersii
G. comptonii
G. monticola
G. recurvus
Outgroup
G. gracilis
Gh Gj
G. ornatus
Gp
Gb
Gm
Gc
Gg
Ga
Gma Gn
G. engysiphon
G. meridionalis
G. caerulus
Long-tongued bee
G. vigilans
G. quadrangulus
Moth
G. trichonemifolius
Short-tongued bee
Sunbird
Long-proboscid fly
Butterfly
G. exilis
G. blomesteinii
Figure 5: Parsimony haplotype network based on a concatenated eight-plastid region matrix (matK, psbA-trnH, trnS-trnG, rpl32-trnL, trnQrps16, petL-psbE, psbJ-petA, ndhJ-trnF) from a sample of 27 species of the Cape radiation clade, plus 3 outgroup species (not shown). Black
circles represent mutational steps; circle size is proportional to frequency; colors indicate pollination system. Gladiolus species: Ga p G.
aureus, Gb p G. brevifolius, Gc p G. carinatus, Gg p G. griseus, Gh p G. hirsutus, Gj p G. jonquilliodorus, Gm p G. martleyi, Gma p
G. maculatus, Gn p G. nerineoides, Gp p G. priorii.
pollination system may have influenced diversification
processes in Gladiolus by directly affecting rates of speciation and/or extinction. Our BiSSE analyses revealed an
important character-dependent effect of pollination system on rates of evolution: species with rarer derived pollination strategies (short-tongued bee, bird, moth, longtongued fly, butterfly, and beetle) had significantly higher
combined rates of speciation and diversification than species with the ancestral long-tongued bee system. To our
knowledge, this is the first group where such an effect is
shown for pollination systems. Recent applications of character-dependent diversification models on southern African and tropical clades found no significant effect of the
type of pollination system or of particular floral traits on
diversification rate (Armbruster et al. 2009; Schnitzler et
al. 2011), whereas a study of Ipomoea did find a strong
character-dependent effect of flower traits on diversifica-
tion but did not specifically test the effect of pollination
systems (Smith et al. 2010). Our finding reveals that certain
types of pollination systems may accelerate species production, suggesting that specific classes of pollination system may have promoted diversification in southern Africa.
In other words, in the case of Gladiolus, species with rarer
pollination systems have overall given rise to new species
more rapidly than species pollinated by the predominant
long-tongued bee pollination system. Our results pose the
new question as to why these rarer pollination systems are
associated with rapid diversification. The answer may lie
in the fact that competition to attract “derived” animal
pollinators is relaxed, since such pollination systems are
less commonly exploited than melittophilous strategies
(Goldblatt and Manning 2006). These rarer, derived systems could be considered key innovations that have effectively opened up new, less crowded pollination niches
Pollination Systems in Gladiolus
Table 4: Results of the Bayes factor (BF) test for different pollination system monophyly hypotheses within the Cape radiation clade
Monophyly constraint
No constraint
Long-tongued bee
pollination
Long-proboscid fly
pollination
Moth pollination
Short-tongued bee
pollination
Sunbird pollination
Marginal likelihood
Ⳳ SE
⫺2log(BF)
⫺10,436.9 Ⳳ .23
⫺10,463.3 Ⳳ .23
52.8
⫺10,437.8 Ⳳ .18
⫺10,452.5 Ⳳ .27
1.9
31.2
⫺10,436.3 Ⳳ .24
⫺10,436.4 Ⳳ .19
1.1
.9
Note: Hypotheses with ⫺2(log BF) 1 10 are rejected. Monophyly of butterfly pollination was not tested given that only one species with that syndrome is found within the Cape radiation clade.
that may have facilitated divergence processes (Hunter
1998).
We suggest that the key for high plant species diversity
in southern African Gladiolus may be the existence of a
diverse array of rare pollinator groups, independently and
in addition to pollination shifts per se. Additional factors
such as variety of soils, ancestral geographic distributions
and climatic conditions should be further explored to account for high rates of species differentiation in one of
the most spectacularly diverse genera of the angiosperms.
Acknowledgments
We thank T. Barraclough, S. Johnson, A. Phillimore, the
members of the Vargas Lab and the Savolainen Lab, and
three anonymous reviewers for comments that improved
the quality of the manuscript; J. Bollback for advice with
the latest version of SIMMAP; and E. Cano for laboratory
assistance. This work was funded by the European Commission (Marie Curie Early Stage Training “Hotspots” network and Marie Curie Intra-European Fellowship
“BirdIsland”).
Literature Cited
Anderson, B., R. Alexandersson, and S. D. Johnson. 2009. Evolution
and coexistence of pollination ecotypes in an African Gladiolus
(Iridaceae). Evolution 64:960–972.
Armbruster, W. S., J. Lee, and B. G. Baldwin. 2009. Macroevolutionary patterns of defense and pollination in Dalechampia vines:
adaptation, exaptation, and evolutionary novelty. Proceedings of
the National Academy of Sciences of the USA 106:18085–18090.
Bakker, F. T., L. W. Chatrou, B. Gravendeel, and P. B. Pelser. 2005.
Plant species-level systematics: new perspectives on pattern and
process. Koenigstein, Koenigstein.
97
Barraclough, T. G. 2006. What can phylogenetics tell us about speciation in the Cape flora? Diversity and Distributions 12:21–26.
Bastida, J. M., J. M. Alcantara, P. J. Rey, P. Vargas, and C. Herrera.
2010. Extended phylogeny of Aquilegia: the biogeographical and
ecological patterns of two simultaneous but contrasting radiations.
Plant Systematics and Evolution 284:171–185.
Bollback, J. 2006. SIMMAP: stochastic character mapping of discrete
traits on phylogenies. BMC Bioinformatics 7:88.
Clement, M., D. Posada, and K. A. Crandall. 2000. TCS: a computer
program to estimate gene genealogies. Molecular Ecology 9:1657–
1659.
Drummond, A. J., and A. Rambaut. 2007. BEAST: Bayesian evolutionary analysis by sampling trees. BMC Evolutionary Biology 7:
214.
Fenster, C. B., W. S. Armbruster, P. Wilson, M. R. Dudash, and J. T.
Thomson. 2004. Pollination syndromes and floral specialization.
Annual Review of Ecology, Evolution, and Systematics 35:375–403.
FitzJohn, R. G., W. P. Maddison, and S. P. Otto. 2009. Estimating
trait-dependent speciation and extinction rates from incompletely
resolved phylogenies. Systematic Biology 58:595–611.
Goldblatt, P. 1996. Gladiolus in tropical Africa. Timber, Portland,
OR.
Goldblatt, P., and J. C. Manning. 1996. Phylogeny and speciation in
Lapeirousia subgenus Lapeirousia (Iridaceae: Ixioideae). Annals of
the Missouri Botanical Garden 83:346–361.
———. 1998. Gladiolus in southern Africa. Fernwood, Cape Town.
———. 2002. Plant diversity of the Cape region of southern Africa.
Annals of the Missouri Botanical Garden 89:281–302.
———. 2006. Radiation of pollination systems in the Iridaceae of
sub-Saharan Africa. Annals of Botany 97:317.
Goldblatt, P., J. C. Manning, and P. Bernhardt. 2001. Radiation of
pollination systems in Gladiolus (Iridaceae: Crocoideae) in southern Africa. Annals of the Missouri Botanical Garden 88:713–734.
Huelsenbeck, J., R. Nielsen, and J. Bollback. 2003. Stochastic mapping
of morphological characters. Systematic Biology 52:131–158.
Hunter, J. P. 1998. Key innovations and the ecology of macroevolution. Trends in Ecology & Evolution 13:31–36.
Johnson, S. D. 1996. Pollination, adaptation and speciation models
in the Cape flora of South Africa. Taxon 45:59–66.
———. 2010. The pollination niche and its role in the diversification
and maintenance of the southern African flora. Philosophical
Transactions of the Royal Society B: Biological Sciences 365:499–
516.
Johnson, S., H. Linder, and K. Steiner. 1998. Phylogeny and radiation
of pollination systems in Disa (Orchidaceae). American Journal
of Botany 85:402.
Johnson, S. D., and H. Kurzweil. 1998. Systematics and phylogeny
of the Satyrium erectum group (Orchidaceae), with descriptions
of two new species from the Karoo region of South Africa. Botanical Journal of the Linnean Society 127:179–194.
Johnson, S. D., and K. E. Steiner. 1997. Long-tongued fly pollination
and evolution of floral spur length in the Disa draconis complex
(Orchidaceae). Evolution 51:45–53.
Jones, C. S., F. T. Bakker, C. D. Schlichting, and A. B. Nicotra. 2009.
Leaf shape evolution in the south African genus Pelargonium
l’ Her. (Geraniaceae). Evolution 63:479–497.
Kay, K. M., P. A. Reeves, R. G. Olmstead, and D. W. Schemske. 2005.
Rapid speciation and the evolution of hummingbird pollination
in Neotropical Costus subgenus Costus (Costaceae): evidence from
98 The American Naturalist
nrDNA ITS and ETS sequences. American Journal of Botany 92:
1899–1910.
Kay, K. M., and R. D. Sargent. 2009. The role of animal pollination
in plant speciation: integrating ecology, geography, and genetics.
Annual Review of Ecology, Evolution, and Systematics 40:637–656.
Kay, K. M., C. Voelckel, J. Y. Yang, K. M. Hufford, D. D. Kaska, and
S. A. Hodges. 2006. Floral characters and species diversification.
Pages 311–325 in L. D. Harder and S. C. H. Barrett, eds. Ecology
and evolution of flowers. Oxford University Press, New York.
Linder, H. P. 2003. The radiation of the Cape flora, southern Africa.
Biological Reviews 78:597–638.
Linder, H. P., S. D. Johnson, M. Kuhlmann, C. A. Matthee, R. Nyffeler, and E. R. Swartz. 2010. Biotic diversity in the southern African winter-rainfall region. Current Opinion in Environmental
Sustainability 1:1–8.
Maddison, W. P., and D. R. Maddison. 2009. Mesquite: a modular
system for evolutionary analysis, version 2.72. http://
mesquiteproject.org.
Maddison, W. P., P. E. Midford, and S. P. Otto. 2007. Estimating a
binary character’s effect on speciation and extinction. Systematic
Biology 56:701–710.
Mittermeier, R. A., P. R. Gil, M. Hoffman, J. Pilgrim, T. Brooks, C.
G. Mittermeier, J. Lamoreux, and G. A. B. Da Fonseca. 2005.
Hotspots revisited: Earth’s biologically richest and most endangered terrestrial ecoregions. Conservation International, Washington, DC.
Müller, K. 2005. Seqstate: primer design and sequence statistics for
phylogenetic DNA datasets. Applied Bioinformatics 4:65–69.
Pagel, M. 1999. Inferring the historical patterns of biological evolution. Nature 401:877–884.
Patterson, T. B., and T. J. Givnish. 2004. Geographic cohesion, chromosomal evolution, parallel adaptive radiations, and consequent
floral adaptations in Calochortus (Calochortaceae): evidence from
a cpDNA sequence phylogeny. New Phytologist 161:253–264.
Pérez, F., M. T. K. Arroyo, R. Medel, and M. A. Hershkovitz. 2006.
Ancestral reconstruction of flower morphology and pollination
systems in Schizanthus (Solanaceae). American Journal of Botany
93:1029–1038.
Perret, M., A. Chautems, R. Spichiger, G. Kite, and V. Savolainen.
2003. Systematics and evolution of trive Sinningieae (Gesneriaceae): evidence from phylogenetic analyses of six plastid DNA
regions and nuclear ncpgs. American Journal of Botany 90:445–
460.
Renner, S. S., L. Beenken, G. W. Grimm, A. Kocyan, R. E. Ricklefs,
and J. Kohn. 2009. The evolution of dioecy, heterodichogamy, and
labile sex expression in Acer. Evolution 61:2701–2719.
Ronquist, F., and J. P. Huelsenbeck. 2003. MrBayes 3: Bayesian phylogenetic inference under mixed models. Bioinformatics 19:1572–
1574.
Rymer, P. D., S. D. Johnson, S. Savolainen. 2011. Pollinator behaviour
and plant speciation: can assortative mating and disruptive selection maintain distinct floral morphs in sympatry? New Phytologist
188:426–436.
Schnitzler, J., T. G. Barraclough, J. S. Boatwright, P. Goldblatt, J. C.
Manning, M. P. Powell, T. Rebelo, and V. Savolainen. 2011. Causes
of plant diversification in the Cape biodiversity hotspot of South
Africa. Systematic Biology 60:343–357.
Simmons, M. P., and H. Ochoterena. 2000. Gaps as characters in
sequence-based phylogenetic analyses. Systematic Biology 49:369–
381.
Smith, S., C. Ane, and D. Baum. 2008. The role of pollinator shifts
in the floral diversification of Iochroma (Solanaceae). Evolution
62:793.
Smith, S. D. 2010. Using phylogenetics to detect pollinator-mediated
floral evolution. New Phytologist 188:354–363.
Smith, S. D., R. E. Miller, S. P. Otto, R. G. FitzJohn, and M. D.
Rausher. 2010. The effects of flower color transitions on diversification rates in morning glories (Ipomoea subg. Quamoclit, Convolvulaceae). Pages 202–226 in M. Long, H. Gu, and Z. Zhou, eds.
Darwin’s heritage today. Higher Education, Beijing.
Stamatakis, A. 2006. RaxML-VI-HPC: maximum likelihood-based
phylogenetic analyses with thousands of taxa and mixed models.
Bioinformatics 22:2688–2690.
Stebbins, G. L. 1970. Adaptive radiation of reproductive characteristics in angiosperms. I. Pollination mechanisms. Annual Review
of Ecology, Evolution and Systematics 1:307–326.
Tripp, E. A., and P. S. Manos. 2008. Is floral specialization an evolutionary dead-end? pollination system transitions in Ruellia
(Acanthaceae). Evolution 62:1712.
Valente, L. M., V. Savolainen, J. Manning, P. Goldblatt, and P. Vargas.
2011. Explaining disparities in species richness between Mediterranean floristic regions: a case study in Gladiolus (Iridaceae).
Global Ecology and Biogeography 20:881–892.
Van Der Niet, T., and S. D. Johnson. 2009. Patterns of plant speciation
in the Cape Floristic Region. Molecular Phylogenetics and Evolution 51:85–93.
Van Der Niet, T., S. D. Johnson. and H. P. Linder. 2006. Macroevolutionary data suggest a role for reinforcement in pollination system shifts. Evolution 60:1596–1601.
Waterman, R. J., M. I. Bidartondo, J. Stofberg, J. K. Combs, G.
Gebauer, V. Savolainen, T. G. Barraclough, and A. Pauw. 2011.
The effects of above- and belowground mutualisms on orchid
speciation and coexistence. American Naturalist 177:E54–E68.
Whittall, J. B., and S. A. Hodges. 2007. Pollinator shifts drive increasingly long nectar spurs in columbine flowers. Nature 447:706.
Associate Editor: Tia-Lynn R. Ashman
Editor: Mark A. McPeek